Ditto-Bench tests GPT-6 Astra for robot control: strong on simple tasks, stuck on physics
jmin__cho · x · 2026-10-09
Researchers built Ditto-Bench to stress-test GPT-6 Astra and MolmoAct2 as robot-control policies. Both impress on simple goal-conditioned tasks but struggle with complex geometric constraints, precise contact and improvised tool use. Systematic ablations explain success and failure modes, and the authors explore new capabilities like in-context learning for robotics.
Related event: Ditto-Bench Tests GPT-6 Astra's Robot Control Skills(4 posts)→
More from Embodied
- RoboStrategy to Ring Nasdaq Closing Bell with Dyna Robotics Humanoid at Founders Forum — Rewkang · 2026-10-09
- New Paper: Native Action-Prior Learning from Videos for World Action Models — _akhaliq · 2026-10-09
- Shenzhen Hotel Robot Delivers Surprise Birthday Cake, But "No Human Involved" Isn't Quite True — EleanorOlcott · 2026-10-09
- Open-Source Models Power Home Robot Tidying for Toddlers in Weeks, Not Years — chris_j_paxton · 2026-10-09
- Sony's quadruped walks on cracked pavement with passive-stability spherical feet — BLUECOW009 · 2026-10-09
- OpenAI's new Decisions API with image input tested on a MuJoCo robotics policy — RexDouglass · 2026-10-09